OpenAI’s new anti-jobs program

Stargate will create jobs. But not for humans.

Sign up | Read more Future Perfect stories | What is Future Perfect? Kelsey Piper is a senior writer for Future Perfect. She writes about science, technology, and progress. You can read more of her work here and follow her on X. Kelsey Piper is a senior writer for Future Perfect. She writes about science, technology, and progress. You can read more of her work here and follow her on X.   Hey readers,   On Tuesday, I was thinking I might write a story about the implications of the Trump administration's repeal of the Biden executive order on AI. (The biggest implication: that labs are no longer asked to report dangerous capabilities to the government, though they may do so anyway.) But then two bigger and more important AI stories dropped: one of them technical, and one of them economic.  Andrew Harnik/Getty Images Stargate is a jobs program — but maybe not for humans The economic story is Stargate. In conjunction with companies like Oracle and Softbank, OpenAI co-founder Sam Altman announced a mind-boggling planned $500 billion investment in "new AI infrastructure for OpenAI" — that is, for data centers and the power plants that will be needed to power them.    People immediately had questions. First, there was Elon Musk's public declaration that "they don't actually have the money," followed by Microsoft CEO Satya Nadella's rejoinder: "I'm good for my $80 billion." (Microsoft, remember, has a large stake in OpenAI.)    Second, some challenged OpenAI's assertion that the program will "create hundreds of thousands of American jobs."    Why? Well, the only plausible way for investors to get their money back on this project is if, as the company has been betting, OpenAI will soon develop AI systems that can do most work humans can do on a computer. Economists are fiercely debating exactly what economic impacts that would have, if it came about, though the creation of hundreds of thousands of jobs doesn't seem like one, at least not over the long term.    Mass automation has happened before, at the start of the Industrial Revolution, and some people sincerely expect that in the long run it'll be a good thing for society. (My take: that really, really depends on whether we have a plan to maintain democratic accountability and adequate oversight, and to share the benefits of the alarming new sci-fi world. Right now, we absolutely don't have that, so I'm not cheering the prospect of being automated.)    But even if you're more excited about automation than I am, "we will replace all office work with AIs" — which is fairly widely understood to be OpenAI's business model — is an absurd plan to spin as a jobs program. But then, a $500 billion investment to eliminate countless jobs probably wouldn't get President Donald Trump's imprimatur, as Stargate has. DeepSeek may have figured out reinforcement on AI feedback   The other huge story of this week was DeepSeek r1, a new release from the Chinese AI startup DeepSeek, that the company advertises as a rival to OpenAI's o1. What makes r1 a big deal is less the economic implications and more the technical ones.    To teach AI systems to give good answers, we rate the answers they give us, and train them to home in on the ones we rate highly. This is "reinforcement learning from human feedback" (RLHF), and it has been the main approach to training modern LLMs since an OpenAI team got it working. (The process is described in this 2019 paper.)    But RLHF is not how we got the ultra superhuman AI games program AlphaZero. That was trained using a different strategy, based on self-play: the AI was able to invent new puzzles for itself, solve them, learn from the solution, and improve from there.   This strategy is particularly useful for teaching a model how to do quickly anything it can do expensively and slowly. AlphaZero could slowly and time-intensively consider lots of different policies, figure out which one is best, and then learn from the best solution. It is this kind of self-play that made it possible for AlphaZero to vastly improve on previous game engines.    So, of course, labs have been trying to figure out something similar for large language models. The basic idea is simple: you let a model consider a question for a long time, potentially using lots of expensive computation. Then you train it on the answer it eventually found, trying to produce a model that can get the same result more cheaply.    But until now, "major labs were not seeming to be having much success with this sort of self-improving RL," machine learning engineer Peter Schmidt-Nielsen wrote in an explanation of DeepSeek r1's technical significance. What has engineers so impressed with (and so alarmed by) r1 is that the team seems to have made significant progress using that technique.    This would mean that AI systems can be taught to rapidly and cheaply do anything they know how to slowly and expensively do — which would make for some of the fast and shocking improvements in capabilities that the world witnessed with AlphaZero, only in areas of the economy far more important than playing games.   One other notable fact here: these advances are coming from a Chinese AI company. Given that US AI companies are not shy about using the threat of Chinese AI dominance to push their interests — and given that there really is a geopolitical race around this technology — that says a lot about how fast China may be catching up. It's still January A lot of people I know are sick of hearing about AI. They're sick of AI slop in their newsfeeds and AI products that are worse than humans but dirt cheap, and they aren't exactly rooting for OpenAI (or anyone else) to become the world's first trillionaires by automating entire industries.   But I think that in 2025, AI is really going to matter — not because of whether these powerful systems get developed, which at this point looks well underway, but for whether society is ready to stand up and insist that it's done responsibly.    When AI systems start acting independently and committing serious crimes (all of the major labs are working on "agents" that can act independently right now), will we hold their creators accountable? If OpenAI makes a laughably low offer to its nonprofit entity in its transition to fully for-profit status, will the government step in to enforce nonprofit law?    A lot of these decisions will be made in 2025, and the stakes are very high. If AI makes you uneasy, that's a lot more reason to demand action than it is a reason to tune out.  —Kelsey Piper, senior writer 📲 Questions? Comments? Tell us what you think! If there is a topic you want us to explain or a story you're curious to learn more about, fill out this form or email us at futureperfect@vox.com.   Meet the new neighbors: 7.5 million chickens and their mountains of manure Joe Gough for Vox To live near a mega factory farm is to be surrounded by a sickening stench and hordes of flies. Senior staff writer Kenny Torella spoke to residents in Malcom, Iowa, to unearth how the meat industry negatively affects people's day-to-day lives, from eroding community connections to dirtying the air. Read his feature here.   More on this topic from Vox: Americans are eating less meat. And more meat. How? 4 reasons why factory farming still exists Big Oil and Big Ag are teaming up to turn cow poop into energy — and profits. The math doesn't add up. Americans are ditching traditional health care for something cheaper — and riskier Paige Vickers/Getty Images The traditional health care system has become so busted that people are willing to bet on themselves and a small group of peers and hope for the best. More and more Americans are turning to "health cost-sharing ministries," traditionally a faith-based alternative for those with religious objections to traditional insurance. But that salvation is a mirage, one that leaves them vulnerable to the same overriding problem that Americans with conventional insurance rightfully complain about: Their health care is still too damn expensive, argues senior correspondent and editor Dylan Scott.   More on this topic from Vox: Why do so many young people suddenly have cancer? American doctors hate the health care system almost as much as you do Have we cured AIDS?   Become a Vox Member   Support our journalism — become a Vox Member and you'll get exclusive access to the newsroom with members-only perks including newsletters, bonus podcasts and videos, and more. Join our community WHAT WE'RE THINKING ABOUT BSIP/UIG Via Getty Images One of my favorite things in life is a headline that immediately catches my attention paired with a story that keeps it, too. I had a taste of that this week with a piece in Allure titled "This Therapist Is Not Who She Seems to Be" (bonus points for the equally eye-catching and spooky illustration). I won't spoil too much, but it's an interesting look at how people can game search engines for profit by using a veil of legitimacy — and how AI can make that process of deception even easier. —Sam Delgado, fellow Rather than watch a set of billionaires sit in a place of prominence behind the new president on Inauguration Day, my wife and I opted to watch an exploration of a very different class structure: Robert Altman's 2001 classic Gosford Park, about the aristocrats and servants at a shooting party held at a 1930s country house in England. The best appreciation I've seen of the film comes from critic Jonathan Rosenbaum, who gets that it's doing something more sophisticated than just mocking the (admittedly horrid) British upper-crust: "the sense that what one gets from Gosford Park is partially a function of what one elects to notice." —Dylan Matthews, senior correspondent   Factory farming critics have been contemplating a tax on meat and dairy for years, and a new paper published in the journal Food Policy looks at what would actually happen if Germany embraced the idea. Researchers found that a €201 tax per ton of Co2-equivalent emissions would reduce animal product consumption 15 to 53 percent, depending on the product, and reduce greenhouse gas emissions from the country's agriculture industry by 22.5 percent. It would also generate €8.2 billion, which could be used to fund more sustainable farming or be given out as a climate dividend at €104 per person. It may seem far off, but Denmark recently passed a modest livestock emission tax, which goes into effect in 2030. —Kenny Torrella, senior reporter   David Lynch meant something special to me. The first film of his that I saw was Inland Empire, screening at my college's film fest festival, and it was unlike anything I had ever experienced — it's still quite unlike any other film ever made. I found it deeply affecting because buried within all of the surreal imagery were powerful emotions, pain and resilience, that were practically pouring out of the screen. After Lynch's death, I read this wonderful essay by Zach Vasquez in Bright Wall/Dark Room, which captured this aspect of his work, the startling sense of empathy that isn't as obvious as his macabre blend of the bizarre and mundane. —Dylan Scott, senior correspondent and editor Are you enjoying the Future Perfect newsletter? Forward it to a friend; they can sign up for it right here. Want more Future Perfect in your inbox? We also have two other free newsletters you can subscribe to. There's Meat/Less, a five-week course on how to eat less meat, and Processing Meat, a biweekly newsletter from Kenny Torrella and Marina Bolotnikova analyzing how the meat and dairy industries shape health, politics, culture, and the environment.   Today's edition was produced by Izzie Ramirez and edited by Bryan Walsh. Access the web version of this newsletter here.   This email was sent to punjabsvera@gmail.com. Manage your email preferences or unsubscribe.   View our Privacy Notice and our Terms of Service.   Vox Media, 1201 Connecticut Ave. NW, Floor 12, Washington, DC 20036. Copyright © 2025. All rights reserved.